	US 11,704,593 B1
	Apparatuses and methods for evaluation of proffered machine intelligence in predictive modelling using cryptographic token staking
Richard Malone Craib, San Francisco, CA (US); Geoffrey Bradway, San Francisco, CA (US); and Alexander Dunn, San Francisco, CA (US)
Assigned to Numerai Inc., San Francisco, CA (US)
Filed by Numerai, Inc., San Francisco, CA (US)
Filed on Sep. 14, 2021, as Appl. No. 17/475,041.
Application 17/475,041 is a continuation of application No. 15/937,227, filed on Mar. 27, 2018, granted, now 11,164,107.
Claims priority of provisional application 62/648,595, filed on Mar. 27, 2018.
Claims priority of provisional application 62/477,235, filed on Mar. 27, 2017.
This patent is subject to a terminal disclaimer.
Int. Cl. G06N 99/00 (2019.01); G06N 20/00 (2019.01); G06N 5/04 (2023.01); H04L 29/08 (2006.01); H04L 67/104 (2022.01)

CPC G06N 20/00 (2019.01) [G06N 5/04 (2013.01); H04L 67/104 (2013.01)]

20 Claims

1. A method, comprising:

receiving, at a processor and from each data source compute node from a plurality of data source compute nodes, an estimate from a plurality of estimates, each estimate from the plurality of estimates being based on a first data;

receiving, at the processor, an indication of a plurality of stakes, each stake from the plurality of stakes being associated with a data source compute node from the plurality of data source compute nodes;

storing, in memory, an indication of a predefined feedback resource associated with a smart contract;

ranking each data source compute node from the plurality of data source compute nodes based on the plurality of stakes, to generate a plurality of ranked data source compute nodes;

calculating an accuracy of each estimate from the plurality of estimates by comparing each estimate from the plurality of estimates to second data;

for a first ranked data source compute node from the plurality of ranked data source compute nodes:

decrementing the predefined feedback resource and assigning a token augmentation to the first ranked data source compute node if the accuracy of the estimate associated with the first ranked data source compute node has a logloss of less than a first predefined threshold; and

for each remaining ranked data source compute node from the plurality of ranked data source compute nodes:

determining a value associated with the predefined feedback resource; and

if the value is greater than zero, decrementing the predefined feedback resource and assigning a token augmentation to that ranked data source compute node if the accuracy of the estimate associated with that ranked data source compute node exceeds a second predefined threshold.